A fast multilevel method for matching stereo images

This work presents a method, which performs edge stereo matching at different levels, from significant edges to less significant ones. At each level, the process starts by selecting the significant edges with respect to their gradient magnitude. The selected edges are then matched in order to obtain the reference edges from which less significant edges are matched in the next level. The matching procedure is based on a voting scheme by using the local and global constraints. The performance of the proposed stereo matching method is evaluated for real-time obstacle detection in front of a vehicle using linear cameras.